The present invention relates to an equalizing method and an equalizer using a reference signal, and more particularly, to an equalizing method and an equalizer using a reference signal for efficiently removing or reducing multipath distortion in a digital signal demodulating system.
The current trend in the television market is to pursue large scale, realistic, and high definition TV. The Japanese are using high definition TV (HDTV) based on multiple subnyquist sampling encoding (MUSE) which is an analog transmitting method, the US has adopted the GA-HDTV system proposed by the Grand Alliance (GA), and some cable TV companies are interested in quadrature amplitude modulation (hereinafter referred to as QAM) as a digital transmission method.
Because the GA-HDTV adopts digital vestigial side band (hereinafter referred to as VSB) modulation, it is referred to as a GA-VSB type receiver. Here, VSB is used as an analog image signal modulation method in a conventional TV, and for digital signal modulation in the GA-HDTV. In the early digital spectrum compatible-TV (hereinafter referred to as DSC-TV) 2-VSB and 4-VSB were adapted using two and four levels, respectively, as its modulation method. However, the GA-HDTV adopts 8-VSB for terrestrial broadcast mode using 8-levels and 16-VSB for high speed cable mode using 16-levels as its modulation method.
The GA has proposed a schematic constitution of a receiver for demodulating such VSB signals, and the characteristics of the proposed receiver are as follows. First, the VSB receiver, different from other demodulators, detects data using only the In-phase channel (I-channel) signals and performs sampling according to a symbol rate. Therefore, the VSB receiver design is more simple than a QAM receiver which uses quadratrure channel (Q-channel) and I-channel at the same time, and the VSB receiver detects data at a relatively slow processing speed because, unlike a fractional rate receiver, it processes data at a symbol rate. Moreover, the proposed VSB receiver adopts coherent detection for demodulating modulated signals by reproducing carriers from the receiver for detecting digital data. Compared with incoherent detection, coherent detection is advantageous in that it can reliably detect data having the same signal-to-noise ratio, but the receiver design becomes more complicated due to the carrier reproduction circuit. Therefore, the proposed VSB receiver adopts two-step coherent phase detection using a frequency and phase locked loop (FPLL) and a phase tracking loop (PTL) to detect the phase of a transmitted signal.
The FPLL uses a pilot signal included in the VSB signal for estimating the phase of a transmitted VSB signal. Such a FPLL can easily be established using a conventional frequency error detecting circuit, and its constitution and efficiency is disclosed in document 1!: 1! R. Citta, "Frequency and Phase Lock Loop", IEEE Trans, on Consumer Electronics, vol, CE-23, no.3, pp.358-365, Aug. 1977.
The output of the FPLL is transmitted to the PTL via a channel equalizer, and the PTL functions to remove residual phase noise, i.e., phase error, from the output signal.
The GA-HDTV receiver constitution is nearly the same as a decision directed carrier recovery (DDCR) receiver disclosed in document 2!: 2! E. A. Lee and D. G. Messerschmitt, Digital Communication, Kluwer Academic Publishers, Boston, Mass., 1988, but different in that it estimates the rotating components of signal points using only I-channel sampled data and compensates phase error values based on the result.
FIG. 1 is a block diagram showing the constitution of a GA-HDTV receiver, which will be used to describe the VSB type GA receiver. This block diagram is disclosed in document 3!: 3! Grand Alliance HDTV System Specification, submitted to the ACATS Technical Subgroup, Feb. 1994.
In FIG. 1. a tuner 102 selects a desired channel signal among the HDTV signals received via an antenna. Generally, the output of the tuner 102 is a modulated HDTV signal having a 44 MHZ central frequency and a 6 MHZ bandwidth. However, the output of the tuner 102 is not superior in its inner filter characteristic and thus it passes not only the 6 MHZ band of HDTV signals but also a portion of the neighboring channel signals.
Because the neighboring channel signals cause interference in the desired channel signal, the output of the tuner 102 passes through a surface acoustic wave (SAW) filter 104 whose band width is precisely 6 MHZ.
An intermediate frequency (hereinafter referred to as IF) amplifier 106 maintains the input signal of an analog-to-digital (A/D) converter 112 at a proper level, and its gain is controlled by an auto gain control signal output from an auto gain control (hereinafter referred to as AGC) circuit 110.
Here, the proper level means 8-levels (.+-.1, .+-.3, .+-.5, .+-.7) to which is added a DC offset of 1.25, i.e., -5.75, -3.75, -1.75, 0.25, 2.25, 4.25, 6.25, 8.25.
The tuner 102 includes a high frequency amplifier (hereinafter referred to as an RF amplifier) therein, and thus if the gain of the IF amplifier is not sufficient, the gain of the RF amplifier is sufficiently regulated to match the signal being amplified according to the AGC signal.
By the way, the restoration of the carrier is performed by an FPLL circuit 108, which traces the pilot signals added to the output signal of the IF amplifier 106 and regulates the local oscillation frequency of the tuner 102 so that the pilot signal frequency is zero Hz. As a result, the FPLL circuit 108 restores the carrier and multiplies the restored carrier by the output of the IF amplifier 106 to demodulate it as a base-band signal.
The A/D converter 112 samples the output of the FPLL circuit 108 according to the symbol clock signal restored by a symbol timing and field sync restorer 114, and converts the sampled data to digital data.
The symbol timing and field sync restorer 114 generates symbol clock pulses for controlling the sampling timing of the A/D converter 112, supplies the field sync reference signals stored therein to an NTSC detector 116, restores the field sync by comparing the field reference signals with the field sync transmitted to each field, generates the operation clock signal for the entire system, and supplies a field sync control signal to an equalizer 120.
Referring to FIG. 2, showing a VSB data frame format of a GA-HDTV, the VSB data frame is constituted of two fields, each field constituted of one field sync segment and 312 data segments, and each data segment is constituted of four symbols for segment sync and 828 symbols for data and forward error correction (FEC) code.
Moreover, the segment sync is inserted in the eight-level digital data stream at the beginning of each data segment. Here, the segment sync constitutes a regular pattern of four symbols having +5, -5, -5, and +5 signal levels, respectively, and the other data is randomly constituted of the eight levels. Because the segment sync is a binary number (binary level) and should ensure its own stability without interfering with the NTSC signals of the co-channel, its level is determined to be .+-.5.
Additionally, a field sync segment, the first segment of each field, contains a field sync signal for indicating the initiation of a field (FIELD SYNC #1, and FIELD SYNC #2) therein.
FIG. 3 shows the VSB data field sync format of a GA-HDTV.
As shown in FIG. 3, the field sync segment is constituted of 832 symbols, the segment sync constitutes the first four symbols thereof, pseudo number PN 511 constitutes the following 511 symbols, three PN 63's constitute the following 189 symbols, and 24 symbols indicating 2-VSB, 4-VSB, 8-VSB, and 16-VSB modes and other information (104 symbols) constitute the remaining 128 symbols.
Here, PN 511, a predetermined signal sequence which is indicated as +5 and -5 levels, is used as a training sequence for equalizing, and the three PN 63 sequences are used as a field discriminating signal because the phase of the second PN 63 is reversed in every other field.
Referring to FIG. 1, the NTSC detector 116 including an NTSC removing filter (hereinafter referred to as NRF) filters out carrier components of the NTSC signals from the output of the A/D converter 112 for preventing the HDTV broadcast from being deteriorated under channel coherent circumstances when the HDTV and NTSC are simultaneously broadcast.
A DC offset remover 118 removes the DC offset which is due to the non-linearity of the pilot signal and the A/D converter 112 from the digital signals converted in the A/D converter 112.
That is, a small digital DC level (1.25) is transmitted while being added to the four symbol data segment sync having signal levels of +5, -5, -5, and +5 and the 828 symbol data having random levels among the eight levels (.+-.1, .+-.3, .+-.5, .+-.7), and which has the same effect as the pilot signal is added to the data signal. Moreover, the input and output of the A/D converter 112 show non-linear characteristics.
Therefore, the DC offset (1.25) due to the pilot signal and the DC offset due to the non-linearity of the A/D converter 112 should be removed so that an HDTV signal having an original signal level can be restored in a receiver. Accordingly, the DC offset remover 118 removes the DC offset by detecting the mean DC offset of the field sync, and subtracting the DC offset detected in the output of the NTSC detector 116 therefrom. Here, the DC offset is calculated by taking an average of the field sync in the unit of a frame because, e.g., the phase of the second PN 63 among the 189 symbols used as a field discriminating signal as shown in FIG. 3 is reversed in every other field and the absolute values thereof are the same.
The equalizer 120 removes the multipath distortion, in other words multipath noise, due to the transmitting signal passing through the transmission channel. Such multipath distortion results in a multipath channel due to wave deflection off a mountain, a group of buildings, or an airplane during terrestrial broadcasting. The multipath distortion causes delayed and attenuated image signals to be overlapped with the original signal, and distorts the frequency characteristics of the HDTV signal as well. A more detailed description of the NTSC detector 116 and the equalizer 120 will be given with reference to FIG. 4.
In the transmitter, a signal is modulated using Read-Solomon (hereinafter referred to as "RS") coding, an interleaving process, and trellis coded modulation (TCM) before transmission to reduce symbol errors generated during transmission. When the decoder receives a signal from the transmitter, a PTL 122 compensates for that portion of the phase error not completely corrected in the FPLL 108 and a channel decoder 124 decodes the signal received from the PTL 122.
The channel decoder 124 performs trellis decoding on the output of the PTL 122, deinterleaves the trellis decoded data, and error correction decodes the deinterleaved data.
The source decoder 126 variable length decodes the error correction decoded data output from the channel decoder 124, inverse quantizes the variable length decoded data, inverse discrete cosine transforms (IDCT) the inverse quantized data, decompresses the compressed data to restore the original data, and displays it on a display (not shown).
FIG. 4 shows a detailed partial block diagram of a conventional HDTV receiver of a GA-VSB type, and the constitution thereof is described in the above document 3!.
During simultaneous broadcasting of the NTSC and HDTV signals, e.g., when ATV (Advanced Television) and DVB (Digital Video Broadcasting) signals are included, the NTSC signal has a regular carrier frequency offset (about 0.89 MHZ) compared to the VSB signal. Thus, considered to be in base band, the NTSC signal is the same as a modulated carrier frequency corresponding to the frequency offset. Here, most of the NTSC signal energy is concentrated on the original DC component, i.e., the modulated carrier. Thus, when the NTSC mixing components pass through the NRF 204 of the NTSC detector 116, the modulated carrier component is removed and thus the NTSC signal effect on the HDTV signal is decreased.
The NTSC detector 116, using the reference signal, compares the field sync reference signal stored in the symbol timing and field sync demodulator 114 with the field sync output from the A/D converter 112, and determines whether the NTSC signal is mixed with the VSB signal using the accumulated values of the squared difference values.
In other words, the NTSC detector 116 is constituted of an original path through a mixer 201, a squaring circuit 202, and an integrator 203 for comparing the field sync reference signal output from the symbol timing and field sync restorer 114 with the field sync output from the A/D converter 112, and calculating the value of the accumulated squared result of the comparison, a path through the NRF 204, the NRF 205, a mixer 206, a squaring circuit 207, and an integrator 208 for comparing between the field sync output from the A/D converter 112 via NRF 204 and the field sync reference signal output from the symbol timing and field sync restorer 114 via the NRF 205, and calculating the value of the accumulated squared result of the comparison, a least error detector 209 for detecting and outputting the NTSC mixing HDTV channel components by comparing between the two path values, and a multiplexer 210 for selecting I-channel symbol data of the A/D converter 112 output via the NRF 204 or I-channel symbol data output directly from the symbol timing and field sync restorer 114 by using the detection signal output from the least error detector 209 as a selection control signal.
Here, the least error detector 209 generates an NRF controlling (NRF CON) signal for indicating whether or not the output of the A/D converter 112 follows the NRF 204 path in the NTSC detector 116, and transmits the NRF CON signal to a memory 216 of the equalizer 120 and a filter coefficient calculator 217.
The equalizer 120 is constituted of an L1 (herein 78) tab filter 211 as a forward transversal filter, an L2 (herein 177) tab filter 212, a subtractor 213 for subtracting the output of the L2 tab filter 212 from the output of the L1 tab filter 211, a slicer 214 for selecting the output level of the subtractor 213 from among the predetermined eight levels (.+-.1, .+-.3, .+-.5, .+-.7), a controlling switch 215 for supplying random data to L2 tab filter 212 during a data segment term and the training sequence to the memory 216 during the field sync segment term, the memory 216 for storing the training sequence inserted into the field sync segment output from the slicer 214 in every field according to the NRF controlling signal (NRF CON) of the least error detector 209, and the filtering coefficient calculator 217 for performing an equalizing process by renewing the filtering coefficients of the L1 tab filter 211 and the L2 tab filter 212 using the training sequence output from the subtractor 213 during the field sync segment term and outputting random data in a previous filtered state during the coefficient renewal term so that high speed tracking of a moving ghost signal can be possible. Such an equalizer is called a decision-feedback equalizer (DFE).
The filter coefficient calculator 217 calculates the filtering coefficients using a least mean square (LMS) algorithm in order to reduce the effect of noise on the training sequence used for coefficient renewal. During the coefficient renewal, a mean value is taken in every frame because the VSB signal has a different training sequence phase in every field. So, in this embodiment, the mean value is taken in a frame unit because in the second PN 63, the phase is reversed in every field of the field sync, as shown in FIG. 3. To use field mean, PN 511 and a first PN63 can be used. The LMS algorithm will be described later.
Additionally, the DC offset remover 118 filters out the DC offset by subtracting the DC offset transmitted from the filtering coefficient calculator 217 from the output of the NTSC detector 116. That is, because the field sync is transmitted with the DC offset (1.25) being added thereto, the filter coefficient calculator 217 obtains the DC offset value by taking the mean value of the first and second field syncs (FIELD SYNC #1, FIELD SYNC #2) and transmits it to the subtractor 118.
The equalizer 120 can use the reference signal transmitted in every field, random data and its error, or both at the same time. The equalizer shown in FIG. 4 uses both, i.e., the reference data and the random data. When the reference data is used, the coefficients of the tab filters 211 and 212 are to be adopted using the LMS algorithm.
Such an LMS algorithm is widely used and the basis thereof is to minimize the mean square error. When the input and output of the equalizer 120 are x(n) and z(n), respectively, the reference signal determining level is d(n), and the filtering coefficient is w.sub.i, EQU z(n)=W.sup.T (n-1)X.sup.T (n) (1) EQU e(n)=d(n)-z(n) (2) EQU W(n)=W(n-1)+2.mu..multidot.e(n)X(n) (3)
where,
X.sup.T (n)=X(n), X(n-1), . . . , X(n-N+1)!, PA1 W.sup.T =W.sub.0, W.sub.1, . . . , W.sub.N-1 ! PA1 i=0, 1, . . . , N-1, .mu. is constant and N is the tab number of the filter. PA1 (a) determining whether a signal input to a high definition TV (HDTV) includes a field sync or not; PA1 (b) storing a training sequence in a memory when the field sync is detected in the step (a); PA1 (c) calculating a filtering coefficient according to a predetermined algorithm; PA1 (d) determining the level of the training sequence stored in the step (b) and calculating a symbol error rate (SER) using the pre-stored reference signal; PA1 (e) transmitting the filtering coefficient obtained in the step (c) to the filter when the SER obtained in the step (d) is lower than a predetermined value; and PA1 (f) repeatedly performing the steps (c) through (f) until a loop variable equals a loop constant, which indicates the number of off-line operations capable of being performed in a field, while increasing the loop variable when the SER obtained in the step (d) is greater than or equal to a predetermined value. PA1 a first memory for storing a training sequence carried by the received signal in a field sync term according to a field sync control signal; PA1 a second memory for storing reference signals for 8-level and 15-level processes; PA1 a first filter for filtering the received signal; PA1 a second filter, connected to an output terminal of the equalizer; PA1 a subtractor for subtracting the output of the second filter from the output of the first filter and outputting an equalized signal; and PA1 filter coefficient calculating means for:
The LMS algorithm operates to minimize Ee.sup.2 (n)! and can be adapted to both an on-line system using input data continuously and an off-line system for storing data of a predetermined term in a memory.
In a conventional equalizer, considering the complicated calculation, hardware load, speed, etc., an off-line system is widely adapted rather than an on-line system.
Such an off-line system is also used in a ghost canceler for an NTSC broadcasting and is presently produced on a commercial scale. This ghost canceler is shown in document 4!: K. B. Kim, J. Oh, M. H. Lee, H. Hwang, and D. I. Song, "A New Ghost Cancellation System for Korean GCR," IEEE Trans. on Broadcasting, vol.40, no.3, pp.132-140, Sep. 1994.
The advantages of the off-line system are that it has a high convergent speed due to the predetermined reference signal, and that it can be constructed using a special hardware H/W clock rate, i.e., a clock which is slower than the system clock, considering the H/W speed.
Such an off-line system needs a method for controlling renewal of the filtering coefficients according to the degree of convergence of the LMS algorithm. A general LMS algorithm adopts a mean square error (MSE) detecting method, and the ghost canceler adopts a method for detecting the remaining peak value of a ghost signal using the characteristic of the auto correlation value of the reference data showing peak.
The GA-VSB type equalizer shown in FIG.4 adopts an off-line system which uses the reference signal. Another reason that the GA-VSB type equalizer uses an off-line system is because it performs an NRF process.
Considering simultaneous broadcasting of HDTV and NTSC signals, the GA-VSB type receiver performs an NRF process using a comb filter in order to reduce the NTSC signal to be mixed with the HDTV signal. Because NRF is a subtraction of two signals having full gains, the signal level increases from 8 levels to 15 levels and the carrier-to-noise (C/N) for the 15 level process is decreased by about 3 dB more than the 8 level process. That is, under the 15 level process, a stop & go (SAG) decision-directed algorithm can not be adopted and equalization is performed according to only the reference signal, which causes the convergent speed to deteriorate.
Here, the stop & go (SAG) decision-directed algorithm is described in document 5!: Giorgio Picchi and Giancarlo Prat: "Blind Equalization and Carrier Recovery Using `Stop-and-Go` Decision-Directed Algorithm" IEEE Trans. on Communications. vol., Com-35, no 9, pp 877-887, Sep. 1987.
Consequently, the off-line equalization driven by the reference signal is needed not only for speedy convergence in the initial state and precise equalization under general conditions but also for a converging operation of the 15-level equalizer under the circumstance where an NTSC signal is mixed with an HDTV signal.